Google 為 Chrome Enterprise 增添 Auto Browse 和 Skills

🔑 增強重點摘要

• •Auto Browse leverages agentic AI frameworks to navigate complex multi-step web interfaces, reducing the need for manual API integrations in legacy enterprise applications.
• •The Skills framework utilizes a low-code interface, allowing IT administrators to deploy custom automation scripts across the organization without requiring deep software engineering expertise.
• •Gemini Summary is integrated directly into the Chrome browser's side panel, utilizing real-time context from the active tab to provide enterprise-grade data privacy by ensuring processed content does not train public foundation models.

🛠️ 技術深入

• •Auto Browse utilizes a proprietary 'Browser-Action' model architecture, which maps natural language intent to DOM-level interactions (click, input, scroll) rather than relying solely on traditional API calls.
• •Skills are implemented as containerized browser extensions that execute within a sandboxed environment, ensuring that custom workflows cannot access sensitive user data outside the defined scope.
• •Gemini Summary employs a RAG (Retrieval-Augmented Generation) pipeline that indexes the DOM tree of the active tab, stripping non-essential UI elements before sending the text content to the Gemini API for summarization.
• •IT security controls include new 'AI-Policy' headers that allow administrators to whitelist specific domains where Auto Browse and Gemini features are permitted to operate.

🔮 前景展望AI analysis grounded in cited sources

⏳ 時間線